Artificial neural network modeling of jatropha oil fueled diesel engine for emission predictions
This pa per deals with ar ti fi cial neu ral net work mod el ing of die sel en gine fu eled with jatropha oil to pre dict the un burned hy dro car bons, smoke, and NO x emis sions. The ex per i men tal data from the lit er a ture have been used as the data base for the pro posed neu ral net work model de vel op ment. For train ing the net works, the in jection tim ing, in jec tor open ing pres sure, plunger di am e ter, and en gine load are used as the in put layer. The out puts are hy dro car
... uts are hy dro car bons, smoke, and NO x emis sions. The feed for ward back prop a ga tion learn ing al go rithms with two hid den lay ers are used in the net works. For each out put a dif fer ent net work is de vel oped with required to pol ogy. The ar ti fi cial neu ral net work mod els for hy dro car bons, smoke, and NO x emis sions gave R 2 val ues of 0.9976, 0.9976, and 0.9984 and mean per cent er rors of smaller than 2. 7603, 4.9524, and 3.1136, re spec tively, for train ing data sets, while the R 2 val ues of 0. 9904, 0.9904, and 0.9942, and mean per cent er rors of smaller than 6.5557, 6.1072, and 4.4682, re spec tively, for test ing data sets. The best lin ear fit of re gres sion to the ar ti fi cial neu ral net work mod els of hydrocarbons, smoke, and NO x emis sions gave the cor re la tion co ef fi cient val ues of 0.98, 0.995, and 0.997, re spec tively.